715
edits
WebDriver/RemoteProtocol/Meetings/2019/12/06: Difference between revisions
Jump to navigation
Jump to search
WebDriver/RemoteProtocol/Meetings/2019/12/06 (view source)
Revision as of 15:23, 6 December 2019
, 6 December 2019→Minutes: fill in
m (→Roster: update) |
(→Minutes: fill in) |
||
| Line 13: | Line 13: | ||
== Minutes == | == Minutes == | ||
Remote meeting - Sant Nicolas day | |||
=== Schedule update (mmucci) === | |||
* Status: '''Green''' - In Development and On Track. | |||
* Team has completed 85% of the entire MVP Backlog which is forecast to be completed on Dec 20. | |||
* In total, there are 2 bugs remaining in the MVP Backlog to be completed - 2 in development, 0 available for selection and 0 contingency. | |||
* Parallel development has begun on bugs from the Reserve Backlog - 6 in development, 24 available for selection and 4 completed. | |||
* No schedule risks present. | |||
* Next phase of the project will be planned on Dec 20. | |||
=== 2020 planning === | |||
==== Summary ==== | |||
* Fission work helpful in many areas | |||
** High priority, blocks work on all APIs that somehow interact with content/subframes | |||
* Specification work | |||
* New APIs/backlog investigations | |||
* Puppeteer upstream work | |||
** Make a nicer user experience | |||
How work can be shared: | |||
* "ato:" spec work, schema validation | |||
* "whimboo:" converting to <code>JsWindowActor</code>s (“Fission”) | |||
* "maja_zf:" puppeteer package interaction with Firefox | |||
* all: continued work on high priority APIs | |||
==== Discussion ==== | |||
Most impactful: | |||
* Being able to download Firefox in puppeteer npm package (as part of installation, like Chrome) | |||
* High-value APIs, informed by feedback from users | |||
* Running Puppeteer unit tests against Firefox Nightly in CI | |||
===== Fission (move to JSWindowActors) ===== | |||
* Which events/methods are hindered by not using <code>JsWindowActor</code>s? | |||
** Anything with frames, subframes | |||
** <code>executionContextCreated</code> only works on top-level doc, not subframes | |||
** Logs | |||
* Likely means major refactoring work | |||
* Doesn’t make sense to file specific bugs now, needs more planning/investigation first, we should be aware this is going to be a sizable chunk of our work next quarter | |||
** Only a few domains that are not affected (e.g. domains that only live in the parent process) | |||
** Known unknowns, unknown unknowns, &c. | |||
** ''digitarald:'' Puppeteer for chromium also broken for their site isolation? Are these issues just for Firefox? | |||
*** Puppeteer papers over frame/browsing context tracking, there should be no difference for users between Chromium and Firefox in that respect | |||
*** No upstream work needed in Puppeteer to support Fission | |||
*** ''whimboo:'' Run Puppeteer unit tests with Fission in our CI to see differences | |||
*** ''AutomatedTester:'' Definitely don't want to change an API to get Fission compatibility, this is something that should only happen in the background | |||
* Figure out what is not blocked by Fission so we can parallelize | |||
* Porting to <code>JsWindowActor</code>s should be done progressively, not as one patch | |||
** Gradually hook up more domains to use <code>JsWindowActors</code>? | |||
* Prioritize team work on <code>JsWindowActors</code> in Berlin | |||
===== Spec work + Schema validation ===== | |||
* Do we get schema validation from rust-cdp working, do we do a new thing for WebDriver bi-di, or both? | |||
* We currently develop against the published API docs that are in sync with Chromium upstream and Puppeteer version used in Gutenberg | |||
** No guarantee that these are in sync and we predict there will be subtle differences | |||
* We ought to target a specific CDP version | |||
** This means importing the schema definition from a specific Chromium revision | |||
===== Which APIs are highest priority off the top of your head? ===== | |||
* Question: what do people want to use Puppeteer for? (AutomatedTester) | |||
** "digitarald:" screenshot diffs, visual regression analysis (came up twice); cross browser support (design system team at uber?) -- want to check that nothing breaks in foundational UI | |||
** "whimboo:" logging available from the browser | |||
*** digitarald hasn't heard this | |||
* "ato:" more nuanced work left to do on logging (DevTools code for synthesizing info for logs is very complex and <code>JsWindowActors</code> useful here too) | |||
* "ato:" web scraping (which aligns well with screenshots), prerequisite is navigation (and therefore also blocked by Fission) | |||
* Not high priority but maybe jumping into DevTools debugger (see what Logmein people say in January) | |||
* AutomatedTester will talk to SauceLabs about other requirements from them beside logging | |||
** Generally useful to get debug logs from anyone who uses Puppetee so we can measure method frequency | |||
Next meeting: Fission triage (what is blocked by Fission) and review updated 2020 planning document. | |||
Keep track of easy things to work on in backlog for potential contributors. There is some low-hanging fruit in the <code>Page</code> domain, for example. | |||
=== Using Puppeteer with Firefox === | |||
Harald asked if we need to coordinate with the Puppeteer maintainers about their upcoming release. Is it usable with Firefox Nightly currently? | |||
* It is likely a frustrating experience to use Puppeteer with Firefox right now because of a number of papercuts, missing APIs, and lack of site isolation support | |||
* David has asked the Chrome team to keep quiet about preliminary Firefox support in Puppeteer for now | |||
* This enables a soft-launch in early 2020, and we should evaluate next year what we can do to make the user experience better | |||
== Actions == | |||
* ato: update the 2020 planning document based on priorities set out in this meeting | |||
* ato: file bug about running Puppeteer unit tests with Fission enabled | |||
== Recording == | == Recording == | ||